Professional Documents
Culture Documents
The isosteric heats of adsorption of the components of a gas mixture are critical variables for design of
adsorbers for gas separation. They can be unambiguously defined by the Gibbsian Surface Excess (GSE)
model of multicomponent adsorption. These variables can be experimentally measured by multicomponent
differential calorimetry (MDC) and directly used to describe nonisothermal behavior of practical adsorbers.
There is no need to make simplified assumptions about the nature and size of the adsorbed phase, as required
Downloaded via KING FAHD UNIV PETROLEUM & MINERALS on October 1, 2018 at 06:34:24 (UTC).
by conventional adsorption thermodynamic models, to define the isosteric heats. Pure gas isosteric heats of
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
adsorption of N2 and CO2 on a pelletized silicalite sample were measured using a MDC and a data analysis
algorithm based on the GSE model. The silicalite sample behaved like a homogeneous adsorbent for weakly
polar N2 adsorption. The presence of polar alumina binder in the silicalite sample introduced significant
heterogeneity for more polar CO2 adsorption.
Introduction a unit amount of the adsorbent.7-9 The bulk gas phase is defined
by its pressure (P), temperature (T), and the mole fraction (yi)
The isosteric heats of adsorption of the components of a gas of component i. The Gibbsian adsorbed phase is defined by an
mixture are key thermodynamic variables for design of practical arbitrarily chosen Gibbs’ interface where the local adsorbate
gas separation processes such as pressure swing and thermal density and compositions are the same as those of the
swing adsorption.1,2 They determine the extents of adsorbent homogeneous bulk gas phase. The extent of adsorption of the
temperature changes within the adsorber during the adsorption ith component of the gas mixture is given by its Gibbsian surface
(exothermic) and desorption (endothermic) steps of the pro-
excess (nm i , mol/g) which can be experimentally measured.
cesses. The adsorbent temperature is a key variable in determin-
They are independent of the location of the Gibbs’ interface as
ing the local adsorption equilibria and kinetics on the adsorbent,
long as it is within the bulk gas phase. Thus, the GSE model
which ultimately govern the separation performance of the
does not require that the exact size, structure, location, composi-
processes.
tion, and the density of the actual adsorbed phase be known.
Although many different types of differential and integral
The pressure and temperature of the Gibbsian adsorbed phase
heats of adsorption have been defined in the adsorption
are the same as those of the bulk gas phase.
literature,3-6,8,9 the differential isosteric heats of adsorption are
found to be most useful for practical design of adsorbers.8,9 They All conventional experimental methods for gas adsorption
can be (a) unambiguously defined using the Gibbsian Surface studies (volumetric, gravimetric, chromatographic, etc.) measure
Excess (GSE) model of multicomponent adsorption system, (b) nmi as functions of P, T, and yi under equilibrium conditions
directly measured calorimetrically or indirectly estimated from (multicomponent equilibria) and nm i as functions of time under
multicomponent adsorption equilibrium data, and (c) conve- transient conditions (multicomponent kinetics). They represent
niently used to describe heat balance using the transient GSE the true experimental variables.8
model of batch or flow adsorbers.8-10 Figure 1b schematically shows the conventional model used
The purpose of this work is to (a) review the thermodynamic for multicomponent gas adsorption equilibrium. In this case,
definitions of multicomponent isosteric heats of adsorption and the interface between the bulk gas phase and the adsorbed phase
their use in process design under the framework of the GSE is located exactly where the local density and composition of
model,7-9 (b) discuss limitations of the conventional definition the adsorbed phase approach those of the bulk gas phase. The
of isosteric heats of adsorption using actual amounts adsorbed adsorbed phase is defined by its actual volume (Va, cm3/g) and
as variables, and (c) demonstrate the use of the GSE model in the extent of adsorption of component i is given by its actual
estimating isosteric heats of adsorption from the data obtained amount adsorbed (nai , mol/g).
by a differential adsorption calorimeter. The total void volumes (bulk gas phase + pores in the
adsorbent) for both systems of parts a and b of Figure 1 are the
Isosteric Heats of Adsorption by GSE Model same (V0, cm3/g). They can be experimentally measured by
helium expansion. The actual adsorbed phase volume of Figure
Figure 1a schematically shows the GSE model for an
1b, on the other hand, cannot be unambiguously estimated
equilibrium multicomponent gas adsorption system containing
without making extraordinary assumptions about its structure,
particularly for the meso-micro porous adsorbents of practical
* Corresponding author.
† Present Address: Chemical Engineering Department, Pennsylvania State interest. Another key point to note is that Va is an unknown
University, College Park, PA. function of nai and T (or P, T, and yi).
10.1021/jp9903817 CCC: $18.00 © 1999 American Chemical Society
Published on Web 07/20/1999
6540 J. Phys. Chem. B, Vol. 103, No. 31, 1999 Sirca et al.
Figure 1. Schematic drawings of various adsorption models: (a) Gibbsian surface excess model; (b) conventional model; (c) model for constant
P, T experiment using a pure gas.
A comparative mass balance for the components of the gas It can be shown from the thermodynamic analysis of the GSE
mixture using the models of parts a and b of Figure 1 yields8 model8,9 that, for an ideal gas mixture, one has
[ ]
gas phase and the actual adsorbed phase (Figure 1b), and δln(Pyi)
xai ()nai /∑nai ) is the mole fraction of component i in the actual qi ) RT2 (4)
δT nm
adsorbed phase (Figure 1b). i
Fyi) for a multicomponent gas and when Fa . F for a pure gas. where U° (cal/g) is the total internal energy of the adsorption
Otherwise, there can be a large difference between experimen- system of Figure 1a. The quantity dQ (cal/g) represents the
differential amount of heat to be removed (or added) to the
tally measured nm a
i and calculated ni values. Furthermore, ni
m
a system for a differential change (dnm i ) in the surface excess of
and ni may be approximately equal to each other for the more m
component i at constant T and (nj*i ). The quantity qdi (cal/mol)
selectively adsorbed component of a gas mixture (xai . yi), but
in eq 2 is called the differential heat of adsorption. Equation 3
the difference between nm a
i and ni can be large for the less
gives the relationship between qi and qdi for an ideal gas
selectively adsorbed components of the same gas mixture even
mixture.
though (Fa . F).8
Equation 4 shows that qi can be estimated by knowing the
The conventional approach to develop adsorption thermody-
differential change in the partial pressure (pi ) Pyi) of
namics is to assume nm i ≈ ni and then follow the model of
a
3,4 component i of a gas mixture due to a differential change in
Figure 1b. This can lead to ambiguous and erratic thermo-
system temperature at constant values of nm i . This can be easily
dynamic and kinetic interpretation of experimental adsorption
data. The problem can be resolved by developing adsorption achieved for a pure gas by measuring adsorption isotherms (nm i
thermodynamics and kinetics based on the GSE model of Figure as functions of P) at different constant temperatures8,9 but the
1a.8,9 use of eq 4 for gas mixtures containing two or more components
The equilibrium GSE model defines the isosteric heat of can be formidable if not an impossible task.8 It will be shown
adsorption of component i (qi, cal/mol) as the negative of the later that direct calorimetric estimation of qi for multicomponent
differential change in total enthalpy (H°, cal/g) of the closed systems is much easier.
system of Figure 1a for a differential change in the surface Equation 3 shows that qdi and qi can be easily interchanged
excess of component i (nm as the design variable of choice. We prefer to use qi because of
i ) at constant system temperature (T)
and constant values of surface excesses of other components its traditional thermodynamic connotation with reference to eq
m 4 and its conventional counterpart (eq 21). The second reason
(nj*i ) of the system8,9
to choose qi is its temperature independence over a large range
[ ]
δH° of temperature.9
- ) +qi (2) The GSE model of transient mass and heat balances for ad-
δnm
i m
T,nj*i (de)sorption of an N-component ideal gas mixture in a packed
Isosteric Heat of Adsorption: Theory and Experiment J. Phys. Chem. B, Vol. 103, No. 31, 1999 6541
adsorbent column yields:9,10 adsorbed and equilibrium gas phases are µa and µ, respectively.
For an ideal gas,11
mass balance of component i
( ) [ ]
µa ) µ ) µ*(T) + RT ln P (11)
δnji δ(Gyi)
)- i ) 1, 2, ....N (5) dµ*
δt z δz t µ* - T ) h* (12)
dT
ni ) Fbnm
i + Fyi (6)
where µ* and h* are, respectively, the chemical potential (molar
heat balance Gibbs free energy) and the molar enthalpy of the pure gas at
CAFb () {
δT
δt z
)-
δ
δz
[GCg(T - T0)] } + Fb∑qi ( ) δnm
i
δt z
+
temperature T and a pressure of 1 atm. They are functions of T
only. The molar gas volume for an ideal gas is ν ) (RT/P) and
( )
t
that for the adsorbed phase is νa ()Va/na).
δP It follows from eqs 11 and 12 that
(7)
δt
( ) ( )
z
δµa δ ln P
µa - T ) h* - RT2 (13)
where the transient variables of eqs 5 and 7 represent the gas- δT na δT na
phase pressure (P, atm), the temperature of gas and adsorbent
(T, K), the gas-phase mole fraction of component i (yi), the gas- Equation 10 can be differentiated with respect to na at constant
phase molar density (F, mol/cm3), the gas-phase molar heat P and T and with respect to T at constant na to obtain
( )
capacity (Cg, cal/mol/K), the molar gas flow rate through the
column based on empty cross section of adsorber (G, mol/cm2/ δGa
) µa (14)
s), the surface excess of component i (nmi , mol/g), and the total δna P,T
( ) ( )
amount of component i in the adsorber (nji, mol/cm3), as
δGa δP
functions of a distance z (cm) in the adsorber column at time t ) -Sa + Va (15)
(s). CA (cal/g K) is the heat capacity of the total adsorption δT na δT na
( ) ( ) [( ) ] ( ) ( )
adsorbed phases. All of the above-described variables can be
δSa δµa δ δP δVa δP
experimentally measured unambiguously. - ) - Va a -
δna P,T δT na δn δT na P,T δna P,T δT na
Equations 5-7 can be simultaneously solved with the
(16)
appropriate multicomponent adsorption equilibria and kinetic
models (GSE framework) in conjunction with the appropriate We now consider a process where a differential amount (dn) of
initial and boundary conditions for the steps of the adsorption pure adsorbate is transferred from the gas phase to the adsorbed
process in order to estimate the separation performance of the phase while keeping the system closed (dn + dna ) 0) and
process.2 Thus, they show the practical usefulness of the isosteric maintaining P and T as constants. This can be achieved by fitting
heats of adsorption defined by eqs 2-4 using the GSE model. the gas phase of Figure 1b with a frictionless piston (as shown
by Figure 1c) and then (a) by differentially moving the piston
so as to decrease the total void volume of the system by dV0
Isosteric Heat of Adsorption by Conventional Model (cm3/g) in order to keep P constant and simultaneously (b) by
removing a differential quantity of heat (dQ, cal/g) from the
We now focus on the adsorption thermodynamics for the thermostated system at T. By applying the energy balance for
conventional model of Figure 1b. Only the adsorption from a a closed system with heat transfer and mechanical work at
pure ideal gas will be considered for simplicity. However, a constant P and T,12 one has
fairly detailed development is given because the published
literature on this subject is often not very rigorous.3,4 dU° ) -dQ - P dV0 constant P, T (17)
The key thermodynamic relationships for the adsorbed phase
Where U° ()U + Ua) is the total internal energy of the closed
described by Figure 1b (pure gas only) may be written as3,8
system of part b or c of Figure 1, U ()nu*) is the total internal
energy of the gas phase, n is the total moles of adsorbate in
dUa ) T dSa - P dVa + µa dna (8) that phase (n + na ) constant), and u* (h* ) u* + RT) is the
molar internal energy of the gas at T.
dHa ) T dSa + Va dP + µa dna (9) The total volume of the gas phase (V ) V0 - Va) of parts a
dG ) -S dT + V dP + µ dn
a a a a a
(10) and b of Figure 1 is given by
nRT RT
where Ua, Va, Sa, Ha ()Ua + PVa), Ga ()Ha - TSa) are, V) ; dV ) - dna constant P, T (18)
P P
respectively, the internal energy, the volume, the entropy, the
enthalpy, and the Gibbs free energy of the total adsorbed phase P dV0 ) P dVa - RT dna constant P, T (19)
containing na moles of a pure adsorbate per unit amount of the
adsorbent. The chemical potentials of the adsorbate in the Equations 8-19 can be combined to get
6542 J. Phys. Chem. B, Vol. 103, No. 31, 1999 Sirca et al.
[ ]
δQ
δna ( ) [ ( ){ ( )
) h* -
δHa
δna
) qj 1 -
Va na δqj
V qj δna P,T
+
( ) }]
P,T P,T
na δVa
+1 (20)
Va δna P,T
qj ) RT2 (δ δTln P) na
(21)
( ) δQ
δna P,T
≈ qj (22)
provided that the terms inside the second bracket on the right-
hand side of eq 20 are relatively small. Figure 2. Sketch of the calorimeter sample cell (a) and the calorimeter
Most published literature on adsorption thermodynamics assembly (b).
derives eqs 21 and 22 by assuming that Va , V, and defines the
variable qj as the isoteric heat of adsorption for a pure ideal (V0, cc/g) and dosage (Vd, cc/g) sides are measured. The sample
gas3,4 because eq 21 shows that qj can be calculated from the side is then saturated with a multicomponent gas mixture at P0,
adsorption isosteres (plots of P against T at constant na). T0, and yi0 and the dosage side is filled with a pure gas i (a
Equation 20 shows that eq 22 is strictly valid only when Va component of the gas mixture of interest) at pressure Pd0 (>P0).
, V and the terms [(δ ln qj/δ ln na)P,T] and [(δ ln Va/δ ln na)P,T] A very small amount (∆Ni) of the pure gas i is then slowly
are small. The first criterion may be valid for adsorption of a introduced into the sample side from the dosage side and the
vapor well below its critical temperature, where the adsorbed valve is closed. The sample side is then allowed to reach a new
phase may be liquid like3. Otherwise, eq 22 may not be correct. equilibrium state. The temperature [T(t)] of the gas entering the
The term (δqj/δna)P,T is (a) equal to zero for an energetically sample side is continuously monitored. The total quantity of
homogeneous adsorbent and (b) relatively small for a weakly heat evolved (consumed) in the calorimeter cell (∆Qi) during
heterogeneous adsorbent.9 The term (δυa/δna)P,T is also expected the entire process is obtained by integration of the calibrated
to be small when the absorbed phase is liquid-like. Thus, the thermopile output (voltage vs time). The change in the partial
second criterion for validity of eq 22 may also be satisfied when pressure of component j of the gas mixture in the sample side
υa , υ and the adsorbent is homogeneous. In contrast, the (∆pji) is measured at the end of the process.
definition of q by the GSE model [eqs 2-4] does not require
any assumptions about the nature of the adsorbed phase, and it The amount of pure gas introduced into the sample side during
can be easily used for practical process design purposes. The the above-described process is
parallel between the forms of eq 4 for a pure gas (yi ) 1) and
eq 21 suggests that qi be called the isosteric heat of adsorption
of component i of a gas mixture. A more precise terminology
for qi is “isoexcess heat of adsorption of component i” for a
∆Ni ) ∫0t*Ṅi dt; Ṅi ) Vd[ ]dFd
dt
(23)
V0
∆Ni ) ∆nm
i + ∆pii (25)
RT0
V0
∆Nj*i ) 0 ) ∆nj*i
m
+ ∆pji (26)
RT0
An energy balance for the sample side of the calorimeter (open
system, ∆Hout ) 0, no shaft work) yields9,12
Figure 3. Test results of blank experiment: (a) plot of thermopile
∆Uoi ) ∆Hin - ∆Qi (27)
output signal (voltage vs time); (b) plot of gas temperature entering
calorimeter cell against experimental time.
where ∆Uoi is the total change in internal energy of the sample
side during the experiment. algorithm for data analysis outlined in the previous section. The
It can be shown by combining eqs 2-4 and 24-27 that for general design concepts of our calorimeter system were similar
an ideal gas9 to those used by Myers, Gorte, and their co-workers.13,14 It
{ }
consisted of two identical calorimeter cells (sample and refer-
∆Qi υ0∆pji qj ence) made from cylindrical stainless steel vessels. The sides
∆Ni
) qi - ∑j ∆N - 1 + hhi i ) 1, 2, ... N and the bottom of the calorimeter cells were covered with
i RT0 custom-made thermopiles manufactured by the International
(28)
Thermal Instrument Company of California. The tops of the
where qi is the isosteric heat of adsorption of component i calorimeter vessels were fitted with Zytel lids and thermal shunts
[defined by eqs 2-4] of an N component gas mixture (i ) 1,2, that minimize heat loss through the tops and redirect any heat
... N) at P0, T0, and y0i (or nm 0 m
i and T ). ni is the equilibrium
flow in that direction to the thermopiles. The sample and
surface excess of component i of the gas mixture at P0, T0, and reference cells were installed inside snugly fit holes cut into a
y0i . large aluminum block to maintain constant base temperature
A total of N experiments must be carried out around the base around the thermopiles. The sample cell was connected to a
saturation conditions of P0, T0, y0i (or T0 and nm gas supply chamber (dosage cell) through a metering valve.
i ) by separately
introducing differential quantities of each pure gas (∆Ni, i ) 1, Several multiport valves and interconnecting lines were used
2, ... N). This will generate N independent equations analogous to (a) evacuate the sample cell and flow the initial saturating
to eq 28 that relate N different unknowns (qi). Consequently, qi gas over the clean adsorbent at given values of P0, T0, and y0i ,
can be estimated for a given set of values of T0, P0, y0i (or T0 and (b) evacuate the dosage cell and filling it with the dosing
and nm gas at Pd0 and T0. The sample side was instrumented with
i ).
For adsorption of a pure ideal gas (component i), eq 28 differential pressure cells and thermocouples for continuous
monitoring and recording of pressure and temperature in these
simplifies to (base conditions of T0, P0, y0i ) 1)
cells. A residual gas analyzer (mass spectrometer) was connected
[( ) ( )]
∆Qi ∆Pi to the sample cell through a leak valve for continuously
- hhi - V0 monitoring the gas composition in the cell. A thermocouple
∆Ni ∆Ni
[ ( )]
q0i ) (29) monitored the temperature of the gas entering the sample cell
V0 ∆Pi from the dosage cell during the experiment. The entire
1- calorimetric assembly was kept inside a constant-temperature
RT ∆Ni
air bath. Figure 2 shows a schematic diagram of the sample
where ∆Pi is the change in the gas-phase pressure in the sample cell. The total amount of adsorbent in the cell was typically
side during the process, and qi0 is the isosteric heat of 8-12 g.
adsorption of pure gas i at P0 and T0 (or T0 and nm0 m0 There were two output signals (voltage against time) from
i ). ni is the
equilibrium surface excess of pure gas i at P0 and T0. the calorimeter assembly during the experiments. They were
We now consider the special case of the above-described produced by (a) the sample cell and (b) the reference cell. The
experiment using a pure ideal gas where (a) the gas is difference between these two outputs was used to estimate the
nonadsorbing or (b) the adsorbent is absent, and where hhi is actual heat liberated (consumed) during the ad(de)sorption
equal to zero. The heat evolved in the sample side of the experiment. The signal from the reference cell was used to
calorimeter (∆Q*) for introduction of (∆N*) moles of pure gas, correct for any spurious heat generated inside the calorimetric
in this case (qi ) 0), is given by cell due to temperature fluctuations in the aluminum block.
Figure 3a shows a typical output signal from the calorimeter.
∆Q* ) V0∆P* (30) The voltage-time output signal was transformed into the rate
of heat production vs time data by electrically generating a
where ∆P* is the change in the gas pressure of the sample side known quantity of heat inside the calorimeter and measuring
during the process. V0 is the helium void volume of the sample the output signal. A PC-based data acquisition/control system
side [(a) with adsorbent and (b) without adsorbent]. was set up to control switch valves and to record all experi-
mental data such as pressures, temperatures, compositions, and
Experimental Measurement of Isosteric Heat by heat fluxes as functions of time during the experiments.
Calorimetry
Examples of Calorimetric Data
We constructed a multicomponent differential calorimeter
(MDC) which utilizes the experimental protocol and the Blank Experiments. We carried out several blank experi-
6544 J. Phys. Chem. B, Vol. 103, No. 31, 1999 Sirca et al.
Figure 4. Plot of the quantity of heat liberated (∆Q*) during the blank
experiments (without adsorbents) against pressure change in the
calorimeter cell (∆P*).
Summary
Multicomponent Gas Mixtures in Microporous Adsorbents. In Surfaces of Engineering Thermodynamics; Prentice Hall: Englewood Cliffs, NJ, 1972;
Nanoparticles in Porous Materials; Schwarz, J. A., Contescu, C., Eds.; pp 98-101.
Marcel and Dekker: New York, 1999; Chapter 19, pp 501-528. (13) Dunne, J. A.; Mariwala, R.; Rao, M. B.; Sircar, S.; Gorte, R. J.;
(10) Sircar, S. J. Chem. Soc., Faraday Trans. 1 1985, 81, 1541. Myers, A. L. Langmuir 1996, 12, 5888.
(11) Prausnitz, J. M. Molecular Thermodynamics of Fluid Phase (14) Parrillo, D. J.; Gorte, R. J. Catal. Lett. 1992, 16, 17.
Equilibria; Prentice Hall: Englewood Cliffs, NJ, 1969. (15) Golden, T. C.; Sircar, S. J. Colloid Interface Sci. 1991, 147, 274.
(12) Balzhiser, R. E.; Samuels, M. R.; Eliassen, J. D. Chemical (16) Rosynek, M. P. J. Phys. Chem. 1975, 79, 1280.